The glucose transporter GLUT8 cycles between intracellular vesicles and the plasma membrane. Like the insulin-responsive glucose transporter GLUT4, GLUT8 is primarily located in intracellular compartments under basal conditions. Whereas translocation of GLUT4 to the plasma membrane is stimulated by insulin, the distribution of GLUT8 is not affected by insulin treatment in adipose cells. However, blocking endocytosis by co-expression of a dominant-negative dynamin GTPase (K44A) or mutation of the N-terminal dileucine (LL12/13) motif in GLUT8 leads to accumulation of the glucose transporter at the cell surface in a variety of different cell types. Yeast two-hybrid analyses and GST pulldown assays reveal that the LL signal constitutes a binding site for the β2-adaptin subunit of the heterotetrameric AP-2 adaptor complex, implicating this motif in targeting of GLUT8 to clathrin-coated vesicles. Moreover, yeast two-hybrid assays provide evidence that the binding site for the LL motif maps to the appendage domain of β2-adaptin. To analyze the biological significance of the LL/β2 interaction, we utilized RNA interference to specifically knockdown AP-2. Our results show that RNAi-mediated targeting of the μ2 subunit leads to cellular depletion of AP-2, but not AP-1 adaptor complexes in HeLa cells. As a consequence, GLUT8 accumulates at the plasma membrane at comparable levels to those observed in K44A-transfected cells. Conversely, the intracellular localization of mutant GLUT8-LL/AA is restored by replacing the LL motif in GLUT8 with the transferrin receptor-derived μ2-adaptin binding motif YTRF, indicating that for endocytosis both AP-2 binding motifs can substitute for each other. Thus, our data demonstrate that recruitment of GLUT8 to the endocytic machinery occurs via direct interaction of the dileucine motif with β2-adaptin, and that endocytosis might be the main site at which GLUT8 is likely to be regulated.
Introduction
For most mammalian cells, glucose is the primary source of energy, and specific glucose transport proteins (GLUTs) are required to catalyze the uptake of glucose into the cell. In addition to their complex tissue distribution, the fourteen known mammalian GLUT proteins differ in their substrate specificity, kinetic properties, subcellular localization, and levels of regulation (Joost and Thorens, 2001). Whereas the expression of some GLUT isoforms is tightly restricted to a certain cell type, many types of cells contain more than one GLUT isoform. In adipose cells, GLUT1 is constantly recycling between the plasma membrane and intracellular vesicles, and is equally distributed between both membrane compartments (Al-Hasani et al., 1999). By contrast, adipose cells also contain GLUT4 and GLUT8 that are efficiently sequestered within intracellular membrane compartments. However, whereas insulin stimulation of adipose cells leads to a rapid and reversible redistribution of GLUT4, and to a lesser extent GLUT1, from intracellular vesicles to the plasma membrane, the hormone has no effect on the subcellular localization of GLUT8 (Lisinski et al., 2001). In fact, the stimulus for translocating GLUT8 to the cell surface remains unknown.
A dileucine motif (LL12/13) in the N-terminus of GLUT8 has been shown to be responsible for the intracellular sequestration of GLUT8 in a variety of different cell types, such as Xenopus oocytes, HEK 293T cells, COS-7 cells, rat adipose cells and mouse neuroblastoma cells (Ibberson et al., 2000; Lisinski et al., 2001; Shin et al., 2004). In these cells, it was demonstrated that mutation of the dileucine motif leads to an increase in cell surface expression. In addition, we have demonstrated previously that blocking the endocytosis by co-expression of a dominant-negative mutant dynamin GTPase (K44A) leads to an accumulation of GLUT8 on the cell surface, indicating that the transporters are constantly recycling between distinct intracellular vesicles and the plasma membrane via a dynamin-dependent pathway (Lisinski et al., 2001). Thus, these data suggest that the N-terminal dileucine motif in GLUT8 might constitute a docking site for the endocytosis machinery.
Dileucine- and tyrosine-based motifs have been implicated as internalization signals for various membrane proteins (Bonifacino and Traub, 2003). It has been proposed that these motifs bind (directly or indirectly) to clathrin-associated adaptor protein complexes (APs), thereby leading to the recruitment of membrane proteins into clathrin-coated pits, and eventually to their incorporation into clathrin-coated vesicles (CCVs) (Schmid, 1997). Of the four known heterotetrameric AP complexes, the AP-2 adaptor plays the central role in CCV formation during endocytosis. AP-2 is composed of four subunits, α, β2, μ2 and σ2. It is now well established that tyrosine-based motifs (consensus sequence YxxØ, where x is any amino acid and Ø is a bulky hydrophobic residue) are recognized by the 50 kDa μ2 subunit (Bonifacino and Traub, 2003). By contrast, recognition of leucine-based signals by adaptors is less well characterized. It has been shown that certain LL motifs may bind to the β subunits in APs (Rapoport et al., 1998; Yao et al., 2002). However, others have reported that LL motifs may also bind to the μ subunits of APs (Hofmann et al., 1999; Rodionov and Bakke, 1998). Moreover, a previous study demonstrated that some LL motifs are bound by a hemicomplex of the γ and σ1 subunits in AP-1, and the δ and σ3 subunits in AP-3 (Janvier et al., 2003). Lastly, the recent discovery of GGAs (Golgi associated, γ-adaptin homologous, ARF binding proteins), a group of clathrin-binding monomeric adaptor proteins that also bind certain LL motifs suggests that recognition of LL-based signals may involve additional proteins (Bonifacino and Traub, 2003).
In the present study, we have examined the role of the dileucine motif in the N-terminus of GLUT8 in HeLa cells transfected with a recombinant GLUT8 that contained an HA-epitope tag in a large exofacial loop. Our results demonstrate that the LL-based sorting motif in GLUT8 is a binding site for the β2-adaptin subunit of the AP-2 clathrin adaptor protein complex. RNAi-mediated depletion of AP-2 in HeLa cells leads to accumulation of HA-GLUT8 at the plasma membrane without affecting the targeting of a mutant (LL/AA) HA-GLUT8.
Results
Expression and subcellular localization of HA-GLUT8 in HeLa cells
We have shown previously that in rat adipose cells as well as in COS-7 cells, ectopically expressed GLUT8 is sequestered in intracellular vesicles (Lisinski et al., 2001). By contrast, a mutant GLUT8 where the N-terminal dileucine motif (LL12/13) is replaced by alanines is predominantly targeted to the plasma membrane (Lisinski et al., 2001). To characterize the targeting potential of this motif in a human cell line, we transfected HeLa cells that do not express endogenous GLUT8 with plasmids for haemagglutinin (HA)-epitope-tagged GLUT8. The HA-GLUT8 constructs contain the HA-epitope tag in their large extracellular loop (Lisinski et al., 2001). Using western blot analyses with a monoclonal antibody against the HA tag, the immunostaining revealed a single band of 42 KDa and the protein expression levels of HA-GLUT8 in HeLa cells was comparable to that in COS-7 cells (data not shown). The subcellular distribution of the HA-GLUT8 constructs was analyzed by confocal laser scanning microscopy of non-permeabilized and permeabilized cells stained with a monoclonal antibody against the extracellular HA-epitope tag. As illustrated in Fig. 1, permeabilized cells expressing HA-GLUT8 showed a punctate staining pattern in which HA-GLUT8 appeared to be distributed in intracellular membranes. However, non-permeabilized cells expressing wild-type HA-GLUT8 showed no HA staining. By contrast, cells expressing HA-GLUT8-LL/AA showed strong cell surface HA staining under both non-permeabilizing and permeabilizing conditions. Similarly, non-permeabilized cells co-expressing HA-GLUT8 and dominant-negative dynamin showed a pronounced HA staining of the cell surface. Lastly, cells expressing only the dominant-negative dynamin mutant showed no HA staining under both non-permeabilizing and permeabilizing conditions.
Cell-surface expression of HA-GLUT constructs
To quantify the relative amount of HA-GLUT8 on the cell surface we measured the expression of the HA-epitope tag in intact cells by performing an antibody binding assay as described in the Materials and Methods. Briefly, HA-GLUT8-transfected HeLa cells were incubated first with a monoclonal antibody against the HA tag, then with a secondary 125I-labeled anti-mouse antibody. The amount of cell-surface-associated radioactivity was then determined in a scintillation counter. In parallel, the relative protein expression levels of the HA-GLUTs were determined by western blotting using an anti-HA antibody. Fig. 2 illustrates the cell surface expression of the HA-GLUT8 constructs normalized to the relative protein expression levels. Compared to HA-GLUT8, the cell surface expression of the dileucine mutant was increased approximately tenfold. Similar to our previous results from rat adipose cells (Lisinski et al., 2001), co-expression of the dominant-negative dynamin mutant K44A results in an approximate sixfold increase in cell surface expression of wild-type HA-GLUT8.
Interaction of the GLUT8 N-terminus with adaptins
The N-terminal dileucine motif in GLUT8 is responsible for the intracellular sequestration of the protein in rat adipose cells, COS-7 cells and HeLa cells. Because inhibition of endocytosis by dominant-negative dynamin resulted in accumulation of the transporter at the cell surface (Figs 1 and 2), we speculated that LL12/13 constitutes an internalisation signal in GLUT8 for clathrin-mediated endocytosis (CME). Recruitment of membrane proteins into clathrin-coated vesicles requires interactions of their cytosolic internalization signals with clathrin at the plasma membrane, which may occur either directly or indirectly via the clathrin-associated adaptor protein complex (AP-2) and/or accessory proteins (Kirchhausen, 1999; Traub, 2003). In order to test whether the N-terminus of GLUT8 interacts with AP-2, we employed the yeast two-hybrid system (Fields and Song, 1989). The entire cytoplasmic N-terminus of GLUT8 (GLUT8-NT) was fused to the GAL4 DNA-binding domain (GAL4-BD) and served as the bait. As prey, we used the adaptin subunits of AP-2 (α, β2, μ2, and σ2) fused to the GAL4 DNA transcription activation domain (GAL4-AD). Yeast (Saccharomyces cerevisiae strain SFY526) were co-transformed with bait and prey, and assayed for reporter gene (lacZ) activation, i.e. β-galactosidase activity. Fig. 3A illustrates that GLUT8-NT specifically binds to β2-adaptin, but not to the other (α, μ2, σ2) subunits of the AP-2 complex. Most importantly, the binding of the N-terminus of GLUT8 to β2-adaptin was completely abolished when the dileucine motif was mutated to alanine (Fig. 3A). Western blot analysis with an antibody against GAL4 binding domain confirmed that both baits were expressed at comparable levels (Fig. 3A). Interestingly, cells expressing the wild-type GLUT8-NT bait alone exhibited a higher β-galactosidase activity than cells expressing the dileucine mutant bait protein, indicating that the apparent mild autoactivation of the GLUT8 N-terminus is independent of co-expressed adaptins (Fig. 3A and data not shown).
We further investigated whether GLUT8-NT binds to all four known β-adaptin subunits (β1-β4) that are found in the heterotetrameric adaptor proteins AP-1 to AP-4 (Boehm and Bonifacino, 2001). Yeast cells were co-transformed with the GLUT8-NT bait and the β-adaptin prey constructs, and assayed for reporter gene activation. As shown in Fig. 3B the activity of β-galactosidase was similar for both β1- and β2-adaptin, whereas no significant activity was observed when β3- and β4-adaptin were used as prey, indicating that GLUT8-NT binds to β1- and β2-but not to β3- and β4-adaptin. To confirm the results from the yeast two-hybrid assays on the protein level, we performed pulldown reactions with immobilized GST fusion proteins of the N-terminus of GLUT8 (GST-GLUT8-NT; see Materials and Methods). As shown in Fig. 3C, the AP-2 complex from HeLa cells binds to wild-type GST-GLUT8-NT, but not the corresponding LL/AA mutant. Furthermore, in vitro translated β1- and β2-adaptin both bind to GST-GLUT8-NT but not to the GST-GLUT8-NT-LL/AA mutant (Fig. 3D).
Recently, it has been reported that dileucine motifs from the HIV-1 protein Nef and the lysosomal membrane protein (LIMP-II) interact with AP-1 and AP-3 through a complex of two adaptin subunits, γ and σ1 in AP-1, and δ and σ3 in AP-3, respectively (Janvier et al., 2003). Thus, we asked whether the N-terminal dileucine motif in GLUT8 binds to an analogous AP-2-derived α/σ2 hemicomplex. Yeast cells expressing GAL4-BD/GLUT8-NT, σ2-adaptin and GAL4-AD/α-adaptin were assayed for reporter gene activation as described in Materials and Methods but no reporter gene activation was observed (data not shown).
Mapping of the interaction between the dileucine motif in GLUT8 and β-adaptin in the yeast two-hybrid system
The two large subunits of AP-1 (γ, β1) and AP-2 (α, β2) can be divided into three structural domains: the N-terminal trunk domain, the proline/glycine rich hinge region, and the C-terminal appendage or ear domain (Kirchhausen, 1999). To determine which domain of β2-adaptin is responsible for the binding of the dileucine motif in GLUT8 we performed yeast two-hybrid analyses using the GLUT8-NT construct as bait and truncation mutants of β2-adaptin fused to GAL4-AD as prey. The `trunk-hinge' construct contained the entire trunk and hinge domain (M1-S726) whereas the `hinge-ear' construct contained the hinge region and the entire ear domain (I588-Asn937) of β2-adaptin (Fig. 3E). As shown in Fig. 3D, GLUT8-NT interacted with the hinge-ear but not with trunk-hinge, indicating that the C-terminal appendage/ear domain in β2-adaptin is responsible for binding of the dileucine motif in GLUT8.
RNAi-mediated knockdown of AP-2 in HeLa cells
To test the biological significance of the interaction between the GLUT8 N-terminus and β2-adaptin in AP-2, we analyzed the subcellular targeting of HA-GLUT8 in AP-2-depleted HeLa cells. Therefore, we generated several pSuper-based plasmid constructs where oligonucleotides corresponding to the human μ2-adaptin cDNA sequence are expressed as short hairpin (sh)RNA under the control of the human H1 promoter (Brummelkamp et al., 2002). The pSuper plasmids were transfected into HeLa cells, and the expression of the adaptins was analyzed by western blots after 4 and 6 days of cell culture. Two of the three RNAi constructs (μ2-1 and μ2-2) had an effect on the cellular levels of the μ2-adaptin (Fig. 4 and data not shown). As shown in Fig. 4, transfection of HeLa cells with the μ2-1 construct resulted in a marked decrease in the cellular levels of μ2-adaptin after 6 days of cell culture, whereas the μ2-3 construct had no effect (data not shown).
As observed in a previous study from Robinsons's group (Motley et al., 2003), cellular depletion of the medium chain μ2-adaptin also resulted in the reduction of the levels of α-adaptin that constitutes one of the two large subunits of AP-2. Moreover, knockdown of μ2-adaptin also resulted in reduced levels in β-adaptin (Fig. 4). By contrast, in multiple experiments none of the constructs tested had a significant effect on the levels of γ-adaptin of AP-1, indicating a specific downregulation of AP-2 in μ2-1- and μ2-2-transfected HeLa cells (Fig. 4 and data not shown). Interestingly, the amount of α-adaptin, β-adaptin and μ2-adaptin returned to normal levels after 8 days in culture (data not shown).
Inhibition of transferrin uptake in AP-2-depleted HeLa cells
We next tested the functional consequences of the AP-2 knockdown by measuring the uptake of transferrin in AP-2-depleted HeLa cells. The transferrin receptor (TfR) is constitutively recycling between the plasma membrane and the endosomal system (Maxfield and McGraw, 2004). A tyrosine-containing targeting motif (Y20TRF) in the cytoplasmic tail of the TfR is known to bind to μ2-adaptin at the plasma membrane, and this interaction is essential for the clathrin-mediated endocytosis of the receptor (Collawn et al., 1993; Ohno et al., 1995). Prevention of the TfR/μ2 interaction by mutation of the Y20TRF motif (Jing et al., 1990), overexpression of binding-deficient μ2-adaptin (Nesterov et al., 1999), or RNAi-mediated knowdown of μ2-adaptin (Motley et al., 2003) effectively slows down the recycling of the TfR. In the experiment, HeLa cells were transfected with the μ2-1 construct and cultured for 6 days. Then, the cells were incubated with 125I-labeled transferrin at 37°C. After 15 minutes the cells were subjected to an acid wash, lysed, and the incorporated radioactivity was measured in the lysate as described in the Materials and Methods. Compared to the control, the uptake of 125I-transferrin was reduced by approximately 50% in μ2-1-transfected cells (Fig. 5). Because our transfection efficiency typically was 50-70% (data not shown), the observed reduction of transferrin uptake corresponds to a considerable inhibition of TfR recycling.
AP-2 knockdown increases cell-surface expression of HA-GLUT8
In order to investigate the effect of AP-2 depletion on the targeting of GLUT8, HeLa cells were co-transfected with plasmids for HA-GLUT8s and RNAi vectors and cultured for 6 days. Then, the subcellular distribution of the HA-GLUT8 constructs was analyzed by confocal laser scanning microscopy of non-permeabilized cells stained with a monoclonal antibody against the extracellular HA-epitope tag. As shown in Fig. 6A, no HA staining was observed in cells expressing HA-GLUT8 and the control RNAi construct. By contrast, a substantial anti-HA staining of the plasma membrane was observed in cells co-expressing HA-GLUT8 and the μ2-1 construct. In fact, the cell surface-associated fluorescence of cells expressing HA-GLUT8 and μ2-1 RNAi was comparable to the signal obtained with cells that expressed the dileucine mutant and the control RNAi (Fig. 6A).
Furthermore, the amount of HA-GLUT8 was examined by western blot using an anti-HA-antibody (Fig. 6B). As shown in the figure, co-expression of the μ2-1 siRNA construct did not affect the expression of HA-GLUT8 protein. Next, the cell surface expression of the HA tag was analyzed by the antibody binding assay. In cells co-transfected with the control RNAi vector, the cell surface expression of HA-GLUT8 was ∼30% that of the mutant HA-GLUT8-LL/AA, demonstrating the intracellular sequestration of GLUT8 in these cells (Fig. 6B). Co-transfection of the HA-GLUT8s and the μ2-1 targeting construct led to a substantial increase in HA-GLUT8, almost matching that of the mutant HA-GLUT8-LL/AA. By contrast, expression of the μ2-1 targeting construct had no statistically significant effect on the cell surface levels of HA-GLUT8-LL/AA.
Replacement of the dileucine motif in GLUT8-LL/AA with the YTRF motif from the TfR restores endocytosis without significantly affecting the subcellular targeting of HA-GLUT8
Since endocytosis of GLUT8 requires recruitment of AP-2 via the β2-adaptin subunit, we investigated whether a tyrosine-containing targeting signal that is recognized by μ2 can replace the LL motif. Therefore, we constructed a mutant HA-GLUT8 where the two leucines were replaced by the four amino acid targeting sequence YTRF derived from the human transferrin receptor (Fig. 7A; see Materials and Methods). We then analyzed the interaction of the corresponding GAL4/GLUT8-NT constructs with adaptins using the yeast two-hybrid system. As shown in Fig. 7B, binding of the N-terminus of GLUT8 to β2-adaptin was abolished when the dileucine motif was mutated to alanine. However, replacement of this motif with the transferrin receptor sequence resulted in selective binding of μ2-adaptin to GLUT8-NT/YTRF. Thus, substitution of the LL motif with the YTRF sequence resulted in corresponding changes in the binding preferences of GLUT8-NT for β2-adaptin and μ2-adaptin of AP-2. Moreover, according to the β-galactosidase activity, the interaction between the dileucine signal and β-adaptin was similar to that of the YTRF signal and μ2-adaptin (Fig. 7B).
We then transfected HeLa cells with the corresponding HA-GLUT8 constructs and analyzed both protein expression levels and cell-surface expression as described in Materials and Methods. As illustrated in Fig. 8, the protein expression level of HA-GLUT8/YTRF in HeLa cells was comparable to that of wild-type HA-GLUT8. However, compared with the dileucine mutant, the cell surface expression of the HA-GLUT8/YTRF was substantially reduced, indicating that insertion of the YTRF motif in GLUT8-LL/AA resulted in increased intracellular sequestration of the transporter.
In order to characterize the subcellular distribution of the HA-GLUT8s in HeLa cells, we performed co-localization studies by confocal laser scanning microscopy. Cells expressing HA-GLUT8s were incubated with Alexa Fluor 633-labelled transferrin for up to 1 hour at 37°C, fixed, permeabilized and stained for the HA epitope as described in Materials and Methods. As depicted in Fig. 9, there was little, if any, overlap between the HA-GLUT8 staining and the transferrin-positive compartment. However, the staining for the mutant HA-GLUT8/YTRF showed no increase in the overlap with labelled transferrin, indicating that a YTRF motif in GLUT8 is not more efficient in directing the transporter into recycling endosomes than the native dileucine signal. We did not detect any significant amounts of HA-GLUT8 in LysoTracker-positive vesicles (Fig. 10), indicating that GLUT8 is not enriched in the late endosomal/lysosomal pathway. Also, minor overlap of the staining for HA-GLUT8 was observed with Bodipy-TR ceramide (Fig. 10), a lipidic probe of the Golgi apparatus. Minor overlap was observed with AP-2 and AP-1 adaptors (Fig. 10), the latter being localized to the TGN and endosomes (Hinners and Tooze, 2003). Nevertheless, in all cases the staining pattern for HA-GLUT8/YTRF was not different from that for HA-GLUT8 (data not shown), implying that in HeLa cells the N-terminal adaptin binding signal in GLUT8 serves mainly as an internalization motif.
Discussion
Because mutation of the N-terminal dileucine motif (LL12/13) in GLUT8 leads to an increase in cell surface expression of the transporter in several different cell types (Ibberson et al., 2000; Lisinski et al., 2001; Shin et al., 2004), we speculated that LL12/13 constitutes a binding site for the endocytosis machinery. The observation that GLUT8 accumulates on the plasma membrane in cells expressing a dominant-negative mutant dynamin GTPase (K44A; Figs 2 and 3) indicated that GLUT8 might be internalized by clathrin-mediated endocytosis. Two classes of dileucine signals have been proposed to play roles in intracellular targeting of proteins. The DxxLL-type signals are found in proteins that recycle between the TGN and endosomes, such as mannose-6 phosphate receptors, and are recognized by the VHS domain of GGAs, a recently described family of ADP-ribosylation factor (ARF)-dependent clathrin adaptors (Bonifacino and Traub, 2003). The other class of dileucine signal, the [D/E]xxxL[L/I] motif has been implicated as an internalisation signal for various membrane proteins. Acidic residues in positions -4 and -5 from the first invariant leucine have been proposed to play roles in endosomal/lysosomal targeting, whereas the second leucine can be replaced by an isoleucine without loss of the targeting potential (Bonifacino and Traub, 2003). In GLUT8, the sequence N-terminal to the dileucine motif (E8TQP11) conforms to the consensus sequence. In fact, it has been shown recently that mutation of the acidic residue E8 in position -4 results in accumulation of GLUT8 on the cell surface (Augustin et al., 2005). Like YxxØ-based internalization signals, [D/E]xxL[L/I] motifs have been found to bind AP complexes in vitro. However, the binding site for dileucine motifs in AP-2 remains controversial. Various [D/E]xxxL[L/I] motifs were found to bind to μ-adaptins in vitro (Hofmann et al., 1999; Rodionov and Bakke, 1998), whereas others were reported to bind to the β subunits in APs. (Rapoport et al., 1998; Yao et al., 2002). Because [D/E]xxxL[L/I] motif-containing peptides did not co-crystallize with the AP-2 adaptor (Collins et al., 2002) it has been speculated that similar to the YxxØ-/μ-adaptin interaction, complex formation of APs and certain LL motifs might require secondary modifications such as phosphorylation (Conner and Schmid, 2002; Ricotta et al., 2002) and/or accessory adaptor proteins (Traub, 2003). Moreover, it has been recently shown that certain [D/E]xxxL[L/I] motifs may also interact with AP-1 and AP-3 through γ/σ1 and δ/σ3 hemicomplexes, respectively (Janvier et al., 2003).
Our yeast two-hybrid assay clearly demonstrates that the 25 amino acid N-terminal tail of GLUT8 specifically interacts with the β2 subunit of AP-2, and that this interaction is completely abolished after disruption of the LL12/13 motif (Fig. 3A). Despite a weak autoactivation activity of the β-galactosidase reporter gene observed for the wild-type GLUT8 construct (Fig. 3A and data not shown), no significant interaction was detected with any of the other AP-2 subunits, i.e. α, μ2 and σ2. Moreover, yeast two-hybrid analyses revealed a dileucine-motif-dependent interaction of GLUT8 with β1- and β2-but not with β3- and β4-adaptin, respectively (Fig. 3B), indicating that in vivo the glucose transporter might interact with both AP-1 and AP-2 but not AP-3 and AP-4 adaptors. Consequently, the dileucine-motif-dependent binding of the N-terminus of GLUT8 to AP-2 and in vitro translated β1- and β2-adaptins, was confirmed by GST pulldown assays (Fig. 3C,D). Using the yeast three-hybrid system, we also tested whether the LL motif in GLUT8 binds to a hemicomplex composed of the α- and σ2-adaptin of AP-2, but were unable to detect such an interaction (data not shown).
The β2-adaptin subunit of AP-2 comprises a large (60-70 kDa) trunk domain, a ∼100-residue hinge region and a smaller (30 kDa) appendage (or ear) domain (Owen et al., 2000). The hinge region contains a clathrin box that mediates binding of AP-2 to clathrin. The trunk domain has been reported to bind several dileucine-containing peptides (Rapoport et al., 1998; Yao et al., 2002), whereas the ear domain contains binding sites for accessory proteins for clathrin-coated vesicle assembly including β-arrestin, ARH, AP180, epsin and eps15 (Laporte et al., 2002; Owen et al., 2000). Our finding that the N-terminus of GLUT8 binds to the hinge-ear but not to the trunk-hinge construct in the two-hybrid system indicates that this LL motif interacts primarily with the appendage domain of β2-adaptin (Fig. 3E), perhaps emphasizing the important roles of both α and β appendage domains as general protein interaction hubs in APs (Praefcke et al., 2004). A recent study found evidence for in vitro binding of certain [D/E]xxxL[L/I] motifs to bacterially produced AP-2 cores that lack the ear domains (Honing et al., 2005). Thus, it might also be speculated that AP-2 might contain two distinct binding sites, perhaps with different affinities and/or specificities for [D/E]xxxL[L/I] motifs.
To further analyze the role of the GLUT8/β2 interaction we generated RNAi constructs that target the AP-2-specific μ2 subunit. After 4-6 days of expression in HeLa cells, we observed a substantial reduction in the total amount of the 50 kDa μ2-adaptin, the 105 kDa β-adaptin, and the 112 kDa α-adaptin subunit of AP-2, respectively (Fig. 4). For AP-1 and AP-3 it has been shown previously that deleting one of the subunits causes a decrease in the stability of the other subunits, leading to a failure of the incomplete complexes to localize to membranes (Meyer et al., 2000; Peden et al., 2002). Likewise, it was reported recently that RNAi-mediated knockdown of μ2-adaptin in HeLa cells reduces the amount of membrane-bound α-adaptin (Motley et al., 2003). However, according to our western blot analyses of whole cell lysates, depletion of μ2 apparently reduces the total amount of α-adaptin and β-adaptin, indicating that subunits of an incomplete AP-2 complex may eventually become degraded. By contrast, the μ2-RNAi construct did not alter the expression of the AP-1-specific 104 kDa γ-subunit, demonstrating the specificity of the μ2 knockdown (Fig. 4).
The transferrin receptor (TfR) recycles between the plasma membrane and the endosomal system (Maxfield and McGraw, 2004), and the interaction between the tyrosine-containing targeting motif Y20TRF and μ2-adaptin has been shown to be essential for the endocytosis of the receptor (Collawn et al., 1993; Ohno et al., 1995). Thus, we used the TfR as a reporter to evaluate the inhibitory potential of our RNAi construct. Subsequently, the uptake of 125I-transferrin was reduced by approximately 50% in μ2-RNAi-transfected HeLa cells (Fig. 5). Hence, when our transfection efficiency of 50-70% is taken into account, AP-2 knockdown results in a considerable inhibition of TfR recycling in HeLa cells. To test the biological significance of the GLUT8/AP-2 interaction we then measured the cell surface expression of HA-GLUT8 in AP-2-deficient HeLa cells by an anti-HA antibody binding assay (Al-Hasani et al., 1998). As a result, depletion of AP-2 led to a substantial increase in HA-GLUT8 on the cell surface, reaching the levels observed for the mutant HA-GLUT8-LL/AA (Fig. 6). Similarly, analyses of the cells by confocal laser scanning microscopy clearly demonstrated that the cell-surface-associated fluorescence of cells expressing HA-GLUT8 and μ2-1 RNAi was comparable to the signal obtained from cells that expressed the dileucine mutant and the control RNAi (Fig. 6). Hence, the internalisation of GLUT8 is dependent on the interaction of the LL12/13 motif with the β2 subunit of AP-2.
Many transmembrane proteins contain YxxØ signals and/or [D/E]xxxL[L/I] motifs, both of which have been implicated to play roles in internalization and subcellular trafficking (Bonifacino and Traub, 2003). In order to evaluate the targeting potential of both types of signals in the context of GLUT8 recycling, we have constructed a mutant glucose transporter where the β2-adaptin-binding dileucine motif was replaced by the μ2-adaptin-binding targeting signal derived from the transferrin receptor (Y20TRF). As expected, yeast two-hybrid analysis confirmed corresponding changes in adaptin binding preferences for wild-type GLUT8 and YTRF mutant (β2-adaptin → μ2-adaptin) without significant changes in reporter gene activity for both constructs, respectively (Fig. 7). As a result, compared to the LL/AA mutant, the intracellular sequestration of GLUT8 was restored after insertion of the YTRF signal (Fig. 8), indicating that for endocytosis both AP-2 binding motifs can substitute for each other. However, co-localization studies demonstrate that the YTRF motif in GLUT8 is not more efficient in directing the transporter into recycling endosomes than the native dileucine signal. In fact, in HeLa cells the intracellular localization of GLUT8 and the YTRF mutant was similar with respect to endosomes, lysosomes and the TGN (Figs 9 and 10, and data not shown). Thus, in accordance with previous studies using chimeric proteins with simple membrane topology, we show that for the endocytosis, the YTRF sequence is both autonomous and transplantable even in the context of a multimembrane-spanning glucose transporter (Traub, 2003). However, since the steady state distribution of GLUT8 does not appear to depend on the type of AP-binding signal in the N-terminal tail (i.e. YxxØ versus [D/E]xxxL[L/I]), we speculate that additional signals are required to target the glucose transporter protein into its subcellular compartment.
Even though the function of the dileucine motif in GLUT8 endocytosis seems to be independent of the cell type, the identity of the GLUT8 storage compartment might only be revealed by studying cells that naturally express this glucose transporter isoform. Since GLUT8 is a high-affinity glucose transporter (Km ∼2 mM) that cycles between intracellular vesicles and the plasma membrane, it is tempting to speculate that an acute stimulus might lead to a reversible translocation of the protein to the cell surface, thus allowing an influx of glucose into the cell (Joost and Thorens, 2001). However, our data suggest that endocytosis might also be a possible site at which GLUT8 is likely to be regulated. However, a previous study in adipose cells indicated that insulin does not control endocytosis, because the hormone had no effect on the cell surface localization of GLUT8. Thus, further studies are required to characterize the intracellular storage compartment of GLUT8, and to identify the stimulus that may lead to a translocation of the transporter to the cell surface.
Materials and Methods
Antibodies, proteins and clones
Anti-HA antibody (HA.11) was from Covance Research Products, Inc. (Berkeley, CA, USA). Biotin-labelled goat-anti-mouse antibody and FITC-conjugated streptavidin were from Jackson ImmunoResearch Laboratories, Inc. (West Grove, PA, USA), antibodies against GAL4-DNA-binding domain and α-tubulin were from Santa Cruz Biotechnology (Santa Cruz, CA, USA), antibodies against α-, β-, μ2- and γ-adaptin were from BD Biosciences (San Diego, CA, USA). Bodipy-TR ceramide, LysoTracker-Red and Alexa Fluor 633-labelled transferrin were from Molecular Probes/Invitrogen (Karlsruhe, Germany). Sheep anti-mouse IgG (F(ab′)2-fragment) and human transferrin were from Sigma-Aldrich (St Louis, MO, USA) and were iodinated with Na125I (Hartmann Analytics, Braunschweig, Germany) using Iodobeads (Pierce, Rockford, IL, USA) according to the manufacturer's instructions. Construction of the expression plasmids for HA-GLUT8s and dynamin K44A has been described previously (Al-Hasani et al., 1998; Lisinski et al., 2001). The pSuper vector was from OligoEngine (Seattle, WA, USA), the pAS2.1 GAL4 DNA-binding domain (GAL4-BD) plasmid and the pBridge vector were from BD Biosciences (Palo Alto, CA, USA). The GAL4 DNA-activation domain (GAL4-AD) plasmids for the adaptins, pACTII-α, pGADT7β1, pGADβ2, pcDNA3-SP6β2, pGADT7β3 pGADT7β4, pACTII-μ2 and pACTII-σ2 were kindly provided by Juan S. Bonifacino and Ruben C. Aguilar (NICHD, NIH, Bethesda, MD, USA) and Volker Haucke (FU Berlin, Germany).
GST pulldown reactions of adaptins
cDNA fragments corresponding to the N-terminus of GLUT8 (M1-R25) were PCR amplified using the expression plasmids for HA-GLUT8s as templates and subcloned into pGEX3X (Amersham-Pharmacia Biosciences, Piscataway, NJ, USA). Primers were GAATTCATGTCTCCCGAGGAC (sense strand) and GGATCCCCGGCGGCCGCGGGG (anti-sense strand). The resulting plasmid constructs, pGEX-NT-wt and pGEX-NT-LL/AA, were verified by DNA sequencing and the corresponding GST fusion proteins were expressed and purified as described previously (Smith and Johnson, 1988). For pulldown of AP-2 complexes, GST fusion proteins (10 μg) were immobilized on 50 μl glutathione-Sepharose 4B beads (Amersham Biosciences), and incubated with cleared HeLa cell lysates in PBS (pH 7.4, 1 mM EDTA, 1 mM DTT, 0.1% Triton X-100, 1 mM PMSF, 1 mM Na3VO4) overnight at 4°C. The beads were then washed with the same buffer and the samples were eluted into SDS loading buffer and separated by SDS-PAGE. After transfer to nitrocellulose membranes, AP-2 was detected by western blotting for α-adaptin (see below). Full-length β1 and β2-adaptins were in vitro translated using TNT Quick Coupled Transcription/Translation System (T7 for β1 and SP6 for β2; Promega, Madison, WI, USA) and [35S]methionine (10 Ci/μl; Hartmann Analytics) according to the manufacturer's instruction. Immobilized GST fusion proteins (15 μg) were incubated at 4°C for 2 hours with 35S-labelled adaptins in 20 mM Tris-HCl; pH 8, 150 mM NaCl, 0.2% Triton X-100. The beads were washed with the same buffer and bound proteins were eluted in SDS loading buffer. After SDS-PAGE, the adaptins were visualized by autoradiography using a Phosphor Imager (Amersham-Pharmacia).
siRNA constructs for knockdown of μ2-adaptin
For RNAi-mediated silencing of human μ2-adaptin (RefSeq accession NM_004068) we designed three pairs of complementary oligonucleotides with flanking 5′ BglII and 3′ HindIII overhangs. The 19mer μ2-specific target sequences (sense strands: μ2-1, GAGGGTATCAAGTATCGTC; μ2-2, GCTTCTGCCTACCAACGAC; and μ2-3, ACTACAGCGACCATGATGT) were separated by a 9-nucleotide spacer (TTCAAGAGA) from the reverse complement of the same 19-nucleotide sequence. The oligonucleotides were cloned into the BglII and HindIII sites of the pSuper vector and the resulting constructs were verified by DNA sequencing.
Construction of HA-GLUT8-LL/YTRF
Site-directed mutagenesis was performed using the polymerase chain reaction (PCR)-based Quick-Change method (Stratagene, La Jolla, CA, USA). Cloned Pfu DNA polymerase (Stratagene) was used for 16 PCR amplification cycles according to the manufacturer's instructions. cDNA of the HA-GLUT8 mutant, cloned into the expression vector pCDNA3, was used as the template, and the `loop-in' mutagenesis primers (sense strand; mismatches underlined) were: 5′-CAGGAGACGCAGCCGTATACGCGTTTCCGGCCACCGGAAG-3′. The integrity of the mutated sequences was verified by automated DNA sequencing. In the resulting N-terminal sequence of the GLUT8 mutant, the two-amino acid dileucine sequence (LL12/13) is replaced by the four-amino acid motif YTRF from the transferrin receptor without changing other residues in GLUT8 (Fig. 7A).
Cell culture, transfection of cells and cell-surface antibody-binding assay
COS-7 cells were cultured in Dulbecco's modified Eagle's medium containing 10% FCS, 1% penicillin-streptomycin at 37°C, 5% CO2, and transfected with Fugene (Roche, Basel, Switzerland) according to the manufacturer's instructions. HeLa cells were cultured in Eagle's minimal essential medium containing 10% FCS, 1% penicillin-streptomycin at 37°C, 5% CO2, and transfected with Lipofectamine2000 (Invitrogen, Karlsruhe, Germany) according to the manufacturers instructions.
For the cell-surface antibody binding assay, six-well plates with transfected HeLa cells were chilled on ice, incubated with the anti-HA antibody (1:1000 in culture medium) for 90 minutes, washed three times with PBS (4°C) and incubated with 125I-labeled IgG (106 cpm/well in culture medium). After 1 hour the cells were washed three times with PBS and lysed in 100 mM NaOH containing 0.1% Triton X-100. Finally, the cell-surface-associated radioactivity was determined in a γ-counter. The antibody binding assays were performed in duplicate or triplicate, and the values obtained for empty vector-transfected cells were subtracted from all other values to correct for non-specific antibody binding.
Transferrin uptake assay
HeLa cells in 6-well plates were washed with buffered medium (MEM, 20 mM Hepes pH 7.4, 37°C) before 125I-transferrin (Tf) was added to 50 nM (∼106 cpm/well). After 15 minutes at 37°C, the culture dishes were placed on ice, stripped with 0.5 M NaCl, 0.5 M sodium acetate, washed three times with 20 mM Hepes, 150 mM NaCl, 1 mM CaCl2, 5 mM KCl, 1 mM MgCl2, pH 7.4 and lysed in 100 mM NaOH containing 0.1% Triton X-100. Finally, the amount of cellular 125I-Tf was determined in a γ-counter. The amount of 125I-Tf in the lysate obtained in presence of 1 mg/ml unlabeled Tf during the uptake assay (less than 1%) was subtracted from all other values to correct for non-specific Tf binding. The uptake assays were performed in duplicate or triplicate.
Immunostaining of cells and confocal microscopy
HeLa cells were grown on 12 mm glass coverslips, and transfected with the plasmids as indicated in the figure legends. After 48-72 hours, the cells were fixed for 20 minutes with 4% paraformaldehyde. When indicated, the cells were then permeabilized with PBS/0.1% Triton X-100 for 20 minutes, and blocked for 1 hour in PBS/0.1% Tween 20 containing 5% normal goat serum (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA, USA). For immunostaining of HA-GLUTs, the cells were incubated for 1 hour with monoclonal anti-HA antibody [1:1000 in antibody diluent S3022 (DakoCytomation, Carpinteria, CA, USA)], followed by incubation with FITC-labelled goat anti-mouse antibody (1:100; 45 minutes) at room temperature.
AP-1 and AP-2 were stained with monoclonal antibodies against γ-adaptin and μ2-adaptin (1:5000 and 1:250; 1 hour), respectively, followed by incubation with Alexa Fluor 546-labelled goat anti-mouse antibody (1:800; 30 minutes). For staining of the Golgi apparatus, cells were incubated prior fixation for 1 hour with Bodipy-TR ceramide (0.45 μg/ml in PBS, 0.1% Triton X-100, 1% BSA) at room temperature. Lysosomes were stained by incubating living cells with LysoTracker-Red (250-400 nM in MEM/FCS; 2 hours); vesicles of the transferrin recycling pathway were stained by incubation of the cells with Alexa Fluor 633-labelled transferrin for 5-60 minutes at 37°C prior to fixation. The stained cells were mounted on coverslips with Fluoromount (DakoCytomation) and the fluorescence was detected with a confocal laser scanning microscope equipped with 488/546/633 nm lasers (Leica DMIRE2 and TCS SP2; Leica microsystems, Bensheim, Germany).
Yeast two-hybrid analyses
DNA fragments corresponding to the region of the N-terminus of the human GLUT8 (M1-R25) were PCR-amplified to contain a 5′ EcoRI site and a 3′ BamHI site using Pfu DNA polymerase. The primers used were (sense strand): 5′-AGTGAATTCATGTCTCCCGAGGACCCC-3′ and (antisense strand): 5′-AGTGGATCCCCGGCGGCCGCGGGG-3′. The resulting PCR fragments (GLUT8-NTs) were cloned into pAS-2.1 in frame with the GAL4-BD and sequenced. The β2-adaptin deletion constructs, trunk-hinge (M1-S726) and hinge-ear (I588-N937) were generated by subcloning restriction fragments into pGADT7 using Kpn21, SwaI and EcoRI sites, respectively, and the sequence was verified by automated DNA sequencing. Saccharomyces cerevisiae strain SFY526 were co-transformed with DNA-BD and DNA-AD plasmids and maintained on synthetic medium lacking tryptophane and leucine (SD Trp-Leu-). Mid-log phase cultures were grown in complete medium (YPD), harvested, and lysed by freeze-thawing. β-Galactosidase assays were carried out using chloro-phenolred-β-D-galactopyranoside (CPRG, Roche), as described previously (Al-Hasani et al., 2002). Yeast three-hybrid analysis of the interactions of GLUT8-NTs and a combination of two AP-2 subunits, α and σ2 hemicomplexes, was performed essentially as described previously (Janvier et al., 2003). Briefly, the GLUT8-NTs and a cDNA fragment for σ2 were subcloned into MCS-1 and MCS-2 of pBridge vector, respectively, and the resulting construct was introduced into pACTII-α containing SFY526. Mid-log phase cultures were grown in synthetic medium lacking tryptophane, leucine and methionine (SD Trp-Leu-Met-), and assayed for β-galactosidase activity using CPRG.
Western blot analyses
For detection of GAL4-BD bait proteins in yeast, mid-log phase cells were harvested and lysed as described previously (Printen and Sprague, Jr, 1994). For detection of HA-GLUTs, transfected HeLa cells were harvested after 48 hours, and total cellular membranes were prepared as described previously (Al-Hasani et al., 1999). For detection of adaptins, transfected HeLa cells were lysed in lysis buffer (20 mM Tris-HCl, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100, 1 mM PMSF, 1 mM glycerol phosphate, 1 mM Na3VO4, pH 7.4), and the lysates were cleared by a brief centrifugation step. Protein concentration was determined using the bicinchoninic acid protein assay kit (Pierce, Rockford, IL, USA). The samples were subjected to SDS-PAGE (20 μg protein/lane) and transferred onto nitrocellulose membranes. The membranes were incubated with the primary antibody (anti-HA, 1:1000; anti-α-adaptin, 1:1000; anti-β-adaptin, 1:1000; anti-μ2-adaptin, 1:250; anti-γ-adaptin, 1:5000; anti-α-tubulin, 1:500; anti-GAL4BD, 1:1000), washed, incubated again with peroxidase-labelled rabbit anti-mouse IgG and developed by ECL (Amersham Biosciences, Buckinghamshire, UK).
Acknowledgements
We thank Margaret Robinson and Geoffrey Holman for helpful discussions and for critically reading the manuscript; and Brigitte Rischke for expert technical assistance. We also thank Juan S. Bonifacino, Ruben C. Aguilar, Edward B. Leof, Mark Wilkes, and Volker Haucke for providing the adaptin constructs. This work was supported in part by a grant from the Deutsche Forschungsgemeinschaft to H.A. (AL452-2).